Detection of Differentially Regulated Genes in Keratinocytesby cDNA Array Hybridization: Hsp27 and Other NovelPlayers in Response to Arti®cial Ultraviolet Radiation

Bernd Becker, Thomas Vogt, Michael Landthaler, and Wilhelm StolzDepartment of Dermatology, University of Regensburg, Regensburg, Germany

cDNA array technology was used to identify novelgenes participating in the ultraviolet response ofcultured human keratinocytes. cDNA arrays repre-senting more than 50,000 different cDNA cloneswere hybridized with complex probes generated bySMART-polymerase chain reaction ampli®cation of150 ng of total RNA extracted 24 h after ultravioletirradiation. Fifty-one clones with differential hybridi-zation signals were detected, representing 19 differ-ent sequences; 10 known genes (seven ultravioletinduced, three ultraviolet suppressed) and furthernine expressed sequence tags of unknown genes. Inseven of 10 genes the data from cDNA arrays probedwith SMART-cDNA could be con®rmed by north-ern blot analysis. The 27 kDa heat shock protein

mRNA was induced. Keratins 6 and 17, markers forthe hyperproliferative status of keratinocytes, wereamong the ultraviolet suppressed genes. The changeof expression pro®le of keratins indicates a differenti-ation towards a phenotype of keratinocytes presentin the suprabasal layers of the epidermis. Thesemechanisms may contribute to the ultraviolet pro-tective function of the epidermis and to the anti-pro-liferative action of ultraviolet in the therapy ofpsoriasis. We also detected an induction of adenylylcyclase associated protein and the suppression ofG(s)a (a stimulating subunit of the trimeric mem-brane bound GTPase). Key words: cDNA-array/ keratin6/ keratin 17. J Invest Dermatol 116:983±988, 2001

Ultraviolet (UV) irradiation of keratinocytes inducesvarious cascades of changes of gene expression,including both soluble paracrine molecules, e.g.,growth factors and cytokines, as well as non-secreted molecules, e.g., adhesion molecules,

transcription factors, and cell cycle regulators (Inohara et al, 1995;Krutmann and Grewe, 1995; Kondo et al, 1997). Detailedknowledge about the regulation of these pleiotropic effects isessential for the understanding of the complex interaction betweenmelanocytes and keratinocytes forming the so-called epidermalmelanin unit and thereby the analysis of the carcinogenic effects ofUV. The epidermal melanin unit provides the intrinsic sunprotection against UVB, which presents the most importantenvironmental hazard to skin. Even a single exposure to UVBradiation may lead to DNA damage, which is the main trigger forcarcinogenesis. Until now differential gene expression could beanalyzed only in a step by step fashion using techniques such asreverse transcriptase PCR, in situ hybridization, or immunohis-

tochemistry, which are not suitable for a simultaneous analysis ofthe complex changes of gene expression that characterize themammalian UV response. With the human genome project, a newera has begun as high-density cDNA arrays became availableallowing the analysis of the expression of thousands of genes at atime. In order to study the complex UV response of keratinocytes,we applied a series of high density cDNA arrays each representing27,648 IMAGE cDNA clones from the human genome project.

In this study we demonstrate the value of this technique byshowing differential expression of 10 known genes and further ninecDNA clones, which represent genes not yet characterized. Apositive correlation of detection of differential expression betweencDNA array and northern blot analysis can be found. We show thatthe gene encoding the heat shock protein 27 kDa (Hsp27) is inducedby UV irradiation, suggesting its protective function against UV-induced apoptosis in keratinocytes. In addition, we demonstrate thatthe new combination of arrays hybridized with probes, which weregenerated by the ampli®cation of cDNA, synthesized by oligo-dT-primed transcription of 150 ng of total RNA (Clontech SMARTsystem), result in reproducible hybridization patterns. This presents anew dimension of applications in medicine as the array analysis ofsmall tissue biopsies from patients with only minute amounts ofRNA available is essential for studying the intricacies of skin cancerevolution. Furthermore, this study con®rms previous estimatesabout the complexity of transcriptional changes upon UVBirradiation (Friedberg et al, 1995; Vogt et al, 1997).

MATERIALS AND METHODS

Cell culture and treatments Primary keratinocytes (CellSystems, StKatharinen, Germany) were grown to 80% con¯uence with keratinocyte

Manuscript received October 1, 1999; revised January 29, 2001;accepted for publication February 15, 2001.

Reprint requests to: Professor Wilhelm Stolz, Department ofDermatology, University of Regensburg, Franz-Josef-Strauss-Allee 11,93053 Regensburg, Germany. Email: wilhelm.stolz@klinik.uni-regensburg.de

Abbrevations: CAP, adenylyl cyclase associated protein; EST, expressedsequence tag; GAPDH, glycerin aldehyde 3-phosphate dehydrogenase;G(s)a, stimulating subunit of the trimeric membrane bound GTPase; hsp,heat shock protein; IMAGE, Integrated Molecular Analysis of Genomesand their Expression; Krt, keratin; MIC-1, macrophage inhibitorycytokine-1; NAPOR, neuroblastoma apoptose-related RNA bindingprotein; RP, ribosomal protein; SG-2, synaptic glycoprotein 2

0022-202X/01/$15.00 ´ Copyright # 2001 by The Society for Investigative Dermatology, Inc.

983

growth medium (keratinocyte growth medium bulletkit, CellSystems) on150 cm2 culture dishes (Falcon, Heidelberg, Germany). For UVB(312 nm) treatment cells were washed once with phosphate-bufferedsaline at 37°C and irradiated through a thin layer of phosphate-bufferedsaline (2 ml) using the Biometra transilluminator (GoÈttingen, Germany)with 0, 50, 100, and 200 mJ per cm2. Phosphate-buffered saline wassubsequently replaced with keratinocyte growth medium and the cellswere incubated for 24 h. The spectrum of the UV illuminator asprovided by the supplier is very similar to standard FS-40 sunlamps(Brown et al, 2000).

RNA extraction Cells were washed once with phosphate-bufferedsaline and lyzed directly in the culture dish containing 18 ml of Trizol(Gibco Lifetechnologies, Karlsruhe, Germany). Total RNA was isolatedfrom the lysate according to manufacturer's instructions. ContaminatingDNA was eliminated by DNAase treatment using the RNase-free DNasekit from Qiagen (Hilden, Germany). RNA concentration wasdetermined photometrically and quality was checked by agarose gelelectrophoresis.

Preparation of complex double-stranded cDNA probes Double-stranded cDNA was synthesized by reverse transcription of 150 ng oftotal RNA and subsequent ampli®cation of the ®rst strand cDNA usingthe SMART system (Clontech, Heidelberg, Germany). In order to get amaximum yield as well as to avoid abundance normalization, the optimalnumber of cycles was determined. Aliquots were taken from thepolymerase chain reaction (PCR) reaction after 15, 18, 21, 26, and 32cycles, and analyzed by agarose gel electrophoresis. In accordance withthe manufacturer's instructions an optimal number of cycles was assumedwhen a moderately strong smear of cDNA ranging from 0.5 to 6 kbwith several bright bands was seen, which also corresponds to abundanttranscripts. A second reaction was prepared and cycled to the optimalnumber of cycles. The double-stranded cDNA was puri®ed with theQiagen PCR puri®cation kit. 100 ng of the double-stranded cDNA waslabeled radioactively with a32P-deoxycytidine triphosphate by randomprimed Klenow fragment synthesis.

CDNA array hybridization The cDNA arrays number 10 and 19from the IMAGE ®lter collection of the Resource Center PrimaryDatabase (RZPD, Berlin, http://www.rzpd.de) were prehybridized inroller bottles at 38°C for 6 h (3 3 sodium citrate/chloride buffer,50 mM Tris±HCl (pH 7.5), 20 mg per ml tRNA, 20 mg per ml boiledsingle stranded salmon sperm DNA, 1 mM ethylenediamine tetraaceticacid, 1 3 Denhardt's solution). Subsequently, the radioactive probe wasadded to a ®nal concentration of 5 3 106 cpm per ml hybridizationsolution. Hybridization was performed under the same conditions for18 h. Nonspeci®cally bound radioactive probe was washed off withbuffer (0.1 3 sodium citrate/chloride buffer, 0.1% sodium dodecylsulfate) at 65°C for 30 min. Filters were exposed for autoradiography for18 h.

IMAGE cDNA arrays Four sets of two different cDNA arrays (nos 10and 19) were obtained from the RZPD. Filter number 10 representedcDNA clones from the library 188 (Soares melanocyte 2NbHM) andlibrary 189 (Soares fetal placenta 2NbHP8-9W). Filter number 19represented cDNA clones from the libraries 116, 363, 385, 361, 371, and382 (Soares ovary tumor NbHOT, Soares subtracted NhHMPu-S1,Soares 8±9 wk fetus Nb2HF8±9w, NCI_CGAP_GCB1 germinal B cell,Soares testis NHT, NCI_CGAP_Pr2 low-grade preneoplastic lesion).Each IMAGE ®lter contained 27,648 Escherichia coli colonies spotted asduplicates harboring plasmids. E. coli bacteria are spotted from 384 wellplates on to the nylon membranes. After growing, the bacteria are lyzedon the membranes and the DNA is cross-linked by UV irradiation on tothe membrane. The bacteria contained cDNA fragments representingparts of known transcripts, sequenced expressed sequence tags (EST) oryet unsequenced cDNA fragments. Detailed information about the ®ltersand libraries is available at http://www.rzpd.de

Quantitation of hybridization signals Hybridization signals on theautoradiographies were densitometrically analyzed using the PC versionof the NIH-Image software, ScionImage (Scion, MD).

Identi®cation of IMAGE clones The co-ordinates of clonesexhibiting at least a 3-fold difference in hybridization signal intensitywere calculated and entered into the online identi®cation form of theRZPD Berlin (http://www.rzpd.de). The detected clones were obtainedfrom the RZPD as stab cultures. Plasmids were prepared from the clonesand sequenced. The sequences were compared with the NCBI databasesusing the basic local alignment search tool (http://www.ncbi.nlm.nih.gov/BLAST).

Northern blot analysis Total RNA (15 mg) was separated byformaldehyde agarose gel electrophoresis [40 mM 3-[N-morpholino]propane sulfonic acid (pH 7.0), 10 mM sodium acetate, 1 mMethylenediamine tetraacetic acid, 2.2 M formaldehyde] and transferredwith 20 3 sodium citrate/chloride buffer (3 M NaCl, 0.3 M sodiumcitrate, pH 7.0) onto nylon membrane (Magna, Micron Separations,Westborough). The RNA was cross-linked to the membrane by UVCirradiation (125 mJ per cm2) using the Biometra transilluminator.Hybridization conditions were the same as described for the arrayhybridization. DNA fragments for labeling were generated by restrictionof the corresponding IMAGE clone plasmids with EcoRV/NotI.Fragments of the plasmids were separated by agarose gel electrophoresisand the gene-speci®c fragment was extracted from the agarose gel usingthe QIAquick Gel extraction kit (Qiagen, Hilden, Germany). Thefollowing clones were used: IMAGp998K11559, IMAGp998F191927,IMAGp998G191927, IMAGp998N041937, IMAGp998F101930,IMAGp998E20578, IMAGp998O02594, IMAGp998C09586,IMAGp998I03560, and IMAGp998I121869 representing the genesmacrophage inhibitory cytokine-1 (MIC-1), G(S)a, Keratin 6, Keratin17, ribosomal protein (RP) S12, RPL11, adenylyl cyclase associatedprotein (CAP), neuroblastoma apoptose-related RNA binding protein(NAPOR), Synaptic glycoprotein 2 (SG-2), Hsp27, respectively.GAPDH (IMAGp998I101860) and b-actin (IMAGp998P211859) servedas controls. The probe (50 ng of a puri®ed DNA fragment) was labeledradioactively with a32P-deoxycytidine triphosphate by random primedKlenow fragment synthesis. Nonspeci®cally bound radioactive probe waswashed off with buffer (0.1 3 sodium citrate/chloride buffer, 0.1%sodium dodecyl sulfate) at 50°C for 30 min. The membranes wereexposed to autoradiography for 18 h. The northern blots were carriedout two times on independently cultured and UV-treated keratinocytes.

RESULTS

Identi®cation of IMAGE clones representing genesresponsive to UVB irradiation In an attempt to identifygenes responsive to varying doses of UV radiation in keratinocytes,cDNA array technology was used allowing us to pro®le theexpression of about 50,000 genes, which were represented ascDNA fragments spotted on one pair of ®lters (nos 10 and 19).Total RNA was prepared from keratinocytes 24 h after cells hadbeen treated with increasing doses of UV (312 nm) (0, 50, 100, and200 mJ per cm2). The radioactive probes for hybridizing the cDNAarrays were prepared using 150 ng of total RNA as template. Foreach UV dose a pair of two cDNA array ®lters (nos 10 and 19) washybridized. Hybridization signals showing different intensities wereanalyzed densitometrically. Only clones displaying at least a 3-folddifference of hybridization signals were taken into consideration.Among the some 27,000 cDNA clones on each ®lter an average ofabout 500 signals was observed, which represented about 2% of allEST spotted. Twenty-one clones on ®lter no. 10 and 30 clones on®lter no. 19 displayed a differential hybridization signal (» 4% and6%). Owing to a high degree of redundancy within the librariesspotted on the ®lters, the 51 clones corresponded to 10 differentknown genes and nine novel EST sequences. As shown in Table Iseven of the identi®ed genes were UV induced, whereas three wereUV suppressed. The factor of induction or suppression wascalculated from the signal ratio of irradiated (200 mJ) vsnonirradiated cells. Eight of the nine EST without any signi®canthomology to any known gene were UV inducible and one wassuppressible (data not shown).

Con®rmation of UV inducible and suppressible genes bynorthern blot analysis Northern blot analysis (Fig 1) was usedfor con®rming the differential gene expression as indicated bydiffering hybridization signals. All northern analysis were carriedout two times from independently cultured and UV-treatedkeratinocytes to demonstrate the reproducibility. The expressionlevels of the genes analyzed, which were normalized according tothe GAPDH signal, are shown in Fig 1(a) for induced genes and inFig 1(b) for suppressed genes (one experiment of two independentnorthern blot analysis is shown). Hsp27 expression wasconcentration-dependently induced on UV irradiation: 2.4-foldat 50, 2.5-fold at 100, and 4.4-fold at 200 mJ per cm2 (Fig 1a).These factors are very similar to those calculated by array analysis

984 BECKER ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

(Table II). Similarly, we analyzed the change of expression of CAPand NAPOR in comparison with GAPDH. The expression of bothgenes showed an increase after UV irradiation, which con®rmedthe data shown by array analysis, although the factors of inductionfor CAP and NAPOR assessed by cDNA array hybridizationdiffered from those obtained by northern blots (59 vs 1.3 and 4 vs1.4, respectively). SG-2 gene expression was relatively constant asrated by northern blot analysis in one experiment, but showed anincrease of a factor of two in the second experiment (Table II).Also by array analysis an induction was observed (factor 5.5,Table II). One of the array signals proved to be falsely positive: thegene for the ribosomal protein S12 (RPS12) was UV induced by afactor of 37 as calculated by array analysis. In contrast, bothnorthern analysis revealed a 2-fold suppression of RPS12 (Fig 1a,Table II). MIC-1 as well as the ribosomal protein L11 (RPL11)were not detectable in northern blot. Only the second northernanalysis showed an increase of a factor of 1.1 for RPL11 expression.We further analyzed the genes in northern blot, which weresuppressed after UV irradiation (Fig 1b). As shown in Fig 1(b), thefactors of the signal ratios of Krt6/GAPDH gene expression did notchange signi®cantly after irradiation with 50 or 100 mJ per cm2

(1.2 and 1.0, respectively) but decreased to less than 0.3 afterirradiation with 200 mJ per cm2 (Fig 1b). Similarly, the factors ofthe expression ratio of coupling protein G(s)a mRNA decreasedafter 50 mJ per cm2 (2.5-fold), 100 mJ per cm2 (1.4-fold), and200 mJ per cm2 (1.8-fold) (Fig 1b). The factor of the ratio of thekeratin 17 (Krt17) gene expression decreased after UV irradiation3.9-fold. These results correlate with the data from array analysis asshown in the summary in Table II. When b-actin was used fornormalization of the northern blots, the factors of induction orrepression were similar indicating the reliability of the comparisonwith GAPDH (data not shown).

DISCUSSION

Detection of differential gene expression upon UVirradiation A cDNA array hybridization technique was used

for studying the complex effects of UV radiation in keratinocytes.From 27,648 different cDNA clones spotted on two IMAGEcDNA arrays, about 2% hybridized with the complex probesgenerated from keratinocytes. de Saizieu et al (1998) assumed that a2-fold change in signal intensity is signi®cant using oligo nucleotidearrays. As the IMAGE cDNA arrays spotted with E. coli coloniescould bear slightly more variation in the amount of cDNA spotted,we considered those clones as signi®cantly induced or repressed,which displayed an at least 3-fold difference in signal intensity.When applying this as a threshold 4% and 6% of the clones showeddifferential hybridization signals on ®lter no. 10 and no. 19,respectively (Table I). Owing to the redundancy of the cDNA®lters these 51 clones represented 10 known genes and nine EST ofnovel genes. The known genes were selected as candidates forveri®cation in northern blot analysis. Meanwhile ®lters are alsoavailable that are spotted with PCR fragments, thereby ensuring ahigher reproducibility; however, the colony ®lters contain about®ve times more cDNA, increasing the chance to detect novelplayers.

Con®rmation of differential expression of selected genes bynorthern blot analysis In order to be able to ensure thereliability of array hybridization and the RNA/cDNA ampli®cationtechniques, we con®rmed the results from the array analysis forseven of 10 genes by northern blot. One gene turned out to befalsely positive and the remaining two could not be found innorthern blot. As shown in Table II the induction (Hsp27, CAP,NAPOR, SG-2) or repression (Krt6, Krt17, G(S)a) detected wasquite consistent within the two methods. In the case of CAP,NAPOR, SG-2, Krt6, and G(s)a the array showed a morepronounced difference than northern blot. These differencesbetween the two methods might be explained by theampli®cation of cDNA for probe preparation, leading to a shiftin the pro®le of the transcribed messages displayed. The expressionof MIC-1 and RPL11 is probably too low, as no signal in northernblot could be detected but the ampli®cation for probe preparationmight have enhanced the sensitivity of array analysis for these

Table I. Detection of differential transcription of genes by array analysis in UVB irradiated keratinocytes

TranscriptIMAGEclone no.a

No. of clonesdetected

0 mJ percm2

50 mJ percm2

100 mJper cm2

200 mJper cm2 Factorb

UV inducedHsp27 IMAGp998L111998 27 4.33

RPS12 IMAGp998F101930 4 37

MIC-1 IMAGp998K11558 2 11

CAP IMAGp998002594 1 59

Synaptic glycoprotein 2 IMAGp9981035602 1 5.5

RPL11 IMAGp998E20578 1 3

NAPOR 1-3 IMAGp998C09586 1 4

UV suppressed 1

Krt6 IMAGp998E221927,IMAGp998E211927

3 < 0.3

Krtl7 IMAGp998N041937 1 0.29

G(s)a IMAGp998F191927 1 0.29

aOnly one representative clone, others submitted to the RZPD database.bRatio between signal of 200 mJ per cm2 and 0 mJ per cm2 as determined densitometrically.

VOL. 116, NO. 6 JUNE 2001 UV RESPONSE IN KERATINOCYTES 985

messages; however, there was a clear discrepancy in the results fromarray and northern blot analysis regarding the RPS12 expression.Bertucci et al (1999) showed that the intensity of hybridizationsignals depends on the amount of target DNA molecules on thearray available for hybridization. As the accuracy of spotting E. coli

colonies and the amount of plasmid DNA produced by the bacteriadetermines the amount of target DNA molecules on each spot,some variations in hybridization signals might occur caused bydeviations in these factors. Especially small differences between theresults of array and northern blot analysis ± as demonstrated forNAPOR, SG-2, Krt6, and G(s)a ± could be explained by differentamounts of target DNA available for hybridization in each spot.Our results indicate that differences in the mRNA expression couldbe detected despite using a RNA ampli®cation protocol. It appearsto be very important that cycling is carefully monitored by agarosegels and overcycling is to be avoided in order to provide probes forhybridization that reliably represent the expression pro®le of thecells analyzed.

Hsp27 is induced upon UV irradiation in keratino-cytes Using cDNA array analysis of primary keratinocytes, wedetected an induction of Hsp27 upon UV irradiation, which wascon®rmed by northern blot analysis (Fig 1a). It was demonstratedearlier, that several Hsp (e.g., Hsp72, Hsp90) were inducible byUVB irradiation (Maytin, 1992). Hsp27 and Hsp70 are thought toprotect cells from apoptosis induced by various stimuli (Jaattela,1999). Hsp27 appears to be also upregulated upon differentiation ofkeratinocytes Trautinger et al (1995) and Nozaki et al (1997) found,that murine Hsp27 is induced, phosphorylated and translocated intothe nucleus upon UVB stimulation. Similarly, an accumulation ofHsp27 occurring with the differentiation-mediated decrease inembryonic stem cell proliferation (Mehlen et al, 1997) could beobserved. According to our data, one could theorize that theincreased Hsp27 expression in keratinocytes upon UV traumaprevents premature apoptosis and loss of protective epidermalkeratinocytic layers. The prevention of apoptosis upon UVBirradiation by expression of Hsp27 may contribute to the well-documented thickening of UVB-irradiated epidermis to protect thestem cells in the stratum basale. The increase of the epidermalthickness and pigmentation of the skin by increased melaninproduction are most important for intrinsic UVB protection(Rucker et al, 1991); however, the functional activity of Hsp27during differentiation depends on its degree of oligomerization,which in turn depends on its dephosphorylation (Mehlen et al,1997). Hsp27 could probably mediate both functions, theprotection against UV-induced apoptosis as well as thedifferentiation, depending on the intracellular activation status ofthe different signaling kinases. Although the detailed mechanismshave to be further elucidated, our ®ndings about Hsp27signi®cantly show the potential of the array technology to sparksubsequent studies of entirely novel mechanisms.

Classi®cation of other novel players in the UV response inkeratinocytes Among 10 UV responsive genes detected bycDNA array analysis seven were induced, whereas three were UVsuppressed. The ribosomal protein L11, a part of the large 60S

Figure 1. Con®rmation of array results by northern analysis.Total RNA was isolated from keratinocytes 24 h after irradiation withincreasing doses of UVB (0 mJ per cm2, 50 mJ per cm2, 100 mJ percm2, 200 mJ per cm2) and 15 mg of total RNA was subjected tonorthern analysis. The blot was hybridized with the indicated probessubsequently. Autoradiographs (lower panels) were scanned andhybridization signals were quantitated by ScionImage software. Thedensitometric data were normalized with the GAPDH signals. The geneinduction (a) and the gene suppression (b) was expressed as percentage ofGAPDH (% of GAPDH, upper panel).

Table II. Comparison of ratios of gene expression changesas determined by array and northern analysis (factors

determined twice)

Gene Factor by arrayFactor by northern(experiment 1/2)

Hsp27 4.33 4.4/2.0CAP 59 1.3/1.2NAPOR 4 1.4/1.2SG-2 5.5 1.0/2.0RPS12 37 0.5/0.4RPL11 3 NDa/1.1MIC-1 11 ND/NDKrt17 0.29 0.3/0.2Krt6 <0.3 0.6/0.6G(s)a 0.29 0.6/0.5

aNot detected.

986 BECKER ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

ribosomal subunit, can be regarded as a protein generally involvedin protein synthesis and may re¯ect a general activation of theprotein metabolism in keratinocytes after UV irradiation.

Two signaling proteins were detected to be differentiallyregulated upon UV irradiation: CAP was induced, whereasG(S)a was suppressed. The trimeric GTPases are signaling subunitsof all 7-transmembrane helix receptors (e.g., steroid hormonereceptors) (LeVine, 1999). In yeast, CAP was shown to beimportant for the RAS-mediated activation of the adenylyl cyclaseand the cytoskeletal rearrangements leading to morphologicchanges of the cells (Gerst et al, 1991; Zelicof et al, 1996). Thereciprocal change of expression of these two molecules in UV-irradiated keratinocytes could re¯ect a shift of hormone-mediatedgrowth of the cells to other activating stimuli mediated via the RASpathway. As RAS activation presents a hallmark of the UV responsein mammalian cells (Devary et al, 1992), our ®ndings are consistentwith the current concepts of the UV triggered pathways.

By array analysis, MIC-1 gene was shown to be induced, whichcould not be con®rmed by northern blot, as no signal could bedetected, indicating a very low expression of the MIC-1 gene.MIC-1 was shown to be expressed in macrophages, and it wasproposed that it may be an autocrine regulatory factor (Bootcov etal, 1997). In the skin, it could prevent an exaggerated in¯ammatoryresponse induced by cytokines released by epithelial cells (Kondo etal, 1993; Boxman et al, 1996; Gaspari et al, 1996).

Two other genes coding for the SG-2 and the RNA-bindingprotein, NAPOR, were also UV induced, which presents evidenceof an UV-dependent regulation of these genes. A functionalcontext of this result is still elusive.

The keratin 6 and 17 genes were UV suppressed, which wascon®rmed by northern blot analysis. These keratins form hetero-dimers and are characteristic for the hyperproliferative state ofkeratinocytes in culture or under pathologic conditions (Galvin etal, 1989). The expression of keratins 6 and 17 is increased insuprabasal keratinocytes in psoriasis (Leigh et al, 1995; McKay andLeigh, 1995) and, recently, keratin 17 was addressed as a majortarget for autoreactive T lymphocytes in psoriatic lesions(Gudmundsdottir et al, 1999). Swensson et al (1998) demonstratedresults indicating a specialized role for these keratins in ridge skinregions as an adaptation to physical stress. The loss of expression ofthese keratins accompanies keratinocyte differentiation towards thesuprabasal phenotype of normal skin (Galvin et al, 1989). Thus,downregulation of these cytokeratins may re¯ect an ongoingdifferentiation from hyperproliferative keratinocytes towards ker-atinocytes of the higher epithelial cell layers in vivo triggered by UVlight, which may explain the therapeutic effects of UVB irradiationin the treatment of psoriasis. All measurements were performed at24 h after exposure. As the genes analyzed may have individualtime courses of upregulation or downregulation after UV, thesingle time point may miss important loci and exaggerate theimportance of others in the analysis of UV regulation.

Taken together, we add several new aspects to the scenario of theUV response in keratinocytes: The induction of Hsp27 as well asthe suppression of Krt6 and Krt17 indicate a UV-inducedphenotype comparable with keratinocytes of the higher levels ofthe epidermis, which may explain the value of UV therapy inhyperproliferative skin disorders (e.g., psoriasis). From the down-regulation of the G(S)a paralleled by the induction of the CAPgene expression we could conclude that UV-induced differenti-ation could be mediated by a switch in signal transduction from 7-transmembrane helix receptors via the trimeric G-proteins to theactivation of the RAS pathway.

Combination of array technology and RNA ampli®ca-tion The ampli®cation of the cDNA by SMART technologysupplied us with a suf®cient amount of DNA for probe preparationensuring signal intensities, which allow the detection of differencesbetween two sources of RNA. The differences detected by arrayanalysis were reliable as we could con®rm the differences in geneexpression by northern blots for seven of 10 genes. In contrast to

protocols used by other groups, we generated probes for arrayhybridization from as little as 150 ng of total RNA (Bertucci et al,1999; Wang et al, 1999). As we were able to reliably probe arrayswith SMART PCR products starting with minor amounts ofRNA, experiments with ex vivo skin samples are now feasible. Thisis important for the veri®cation of tissue culture studies and makesarray technology applicable for studying skin cancer evolution orother disorders, such as psoriasis, directly in the human system. Abetter understanding of UV responses may help not only toimprove the selection of patients for therapy but also to developUV-protective pharmacons and to elucidate carcinogenesis further.Our results provide the assurance that array technology incombination with SMART-generated probes for hybridizationcan be used as a rapid tool with acceptable accuracy in order toexamine global changes of gene expression upon a particularstimulus.

This study was supported in part by the primary health insurance companies in

Bavaria, by the Wilhelm-Sander-Stiftung 97.21.1/2 and by the Deutsche

Forschungsgesellschaft Vo 416/3.1.

REFERENCES

Bertucci F, Bernard K, Loriod B, et al: Sensitivity issues in DNA array-basedexpression measurements and performance of nylon microarrays for smallsamples. Hum Mol Genet 8:1715±1722, 1999

Bootcov MR, Bauskin AR, Valenzuela SM, et al: MIC-1, a novel macrophageinhibitory cytokine, is a divergent member of the TGF-beta superfamily. ProcNatl Acad Sci USA 94:11514±11519, 1997

Boxman IL, Ruwhof C, Boerman OC, Lowik CW, Ponec M: Role of ®broblasts inthe regulation of proin¯ammatory interleukin IL-1, IL-6 and IL-8 levelsinduced by keratinocyte-derived IL-1. Arch Dermatol Res 288:391±398, 1996

Brown DB, Peritz AE, Mitchell DL, Chiarello S, Uitto J, Gasparro FP: Common¯uorescent sunlamps are an inappropriate substitute for sunlight. PhotochemPhotobiol 72:340±344, 2000

Devary Y, Gottlieb RA, Smeal T, Karin M: The mammalian ultraviolet response istriggered by activation of Src tyrosine kinases. Cell 71:1081±1091, 1992

Friedberg E, Walker G, Siede W (eds). Regulatory Responses to DNA-Damaging Agentsin Eukaryotic Cells. Washington: ASM Press, 1995

Galvin S, Loomis C, Manabe M, Dhouailly D, Sun TT: The major pathways ofkeratinocyte differentiation as de®ned by keratin expression: an overview. AdvDermatol 4:277±299, 1989

Gaspari AA, Sempowski GD, Chess P, Gish J, Phipps RP: Human epidermalkeratinocytes are induced to secrete interleukin-6 and co-stimulate Tlymphocyte proliferation by a CD40-dependent mechanism. Eur J Immunol26:1371±1377, 1996

Gerst JE, Ferguson K, Vojtek A, Wigler M, Field J: CAP is a bifunctional componentof the Saccharomyces cerevisiae adenylyl cyclase complex. Mol Cell Biol11:1248±1257, 1991

Gudmundsdottir AS, Sigmundsdottir H, Sigurgeirsson B, Good MF, ValdimarssonH, Jonsdottir I: Is an epitope on keratin 17 a major target for autoreactive Tlymphocytes in psoriasis? [In Process Citation]. Clin Exp Immunol 117:580±586,1999

Inohara S, Kitano Y, Kitagawa K: Cell cycle regulators in the keratinocyte (cyclin-cdk). Exp Dermatol 4:1±8, 1995

Jaattela M: Escaping cell death: survival proteins in cancer. Exp Cell Res 248:30±43,1999

Kondo S, Kono T, Sauder DN, McKenzie RC: IL-8 gene expression and productionin human keratinocytes and their modulation by UVB. J Invest Dermatol101:690±694, 1993

Kondo S, Kooshesh F, Sauder DN: Penetration of keratinocyte-derived cytokinesinto basement membrane. J Cell Physiol 171:190±195, 1997

Krutmann J, Grewe M: Involvement of cytokines, DNA damage, and reactiveoxygen intermediates in ultraviolet radiation-induced modulation ofintercellular adhesion molecule-1 expression. J Invest Dermatol 105:67S±70S,1995

Leigh IM, Navsaria H, Purkis PE, McKay IA, Bowden PE, Riddle PN: Keratins(K16 and K17) as markers of keratinocyte hyperproliferation in psoriasis in vivoand in vitro. Br J Dermatol 133:501±511, 1995

LeVine H, 3rd: Structural features of heterotrimeric G-protein-coupled receptors andtheir modulatory proteins. Mol Neurobiol 19:111±149, 1999

Maytin EV: Differential effects of heat shock and UVB light upon stress proteinexpression in epidermal keratinocytes. J Biol Chem 267:23189±23196, 1992

McKay IA, Leigh IM: Altered keratinocyte growth and differentiation in psoriasis.Clin Dermatol 13:105±114, 1995

Mehlen P, Mehlen A, Godet J, Arrigo AP: Hsp27 as a switch between differentiationand apoptosis in murine embryonic stem cells. J Biol Chem 272:31657±31665,1997

Nozaki J, Takehana M, Kobayashi S: UVB irradiation induces changes in cellular

VOL. 116, NO. 6 JUNE 2001 UV RESPONSE IN KERATINOCYTES 987

localization and phosphorylation of mouse HSP27. Photochem Photobiol65:843±848, 1997

Rucker BU, Haberle M, Koch HU, Bocionek P, Schriever KH, Hornstein OP:Ultraviolet light hardening in polymorphous light eruptionÐa controlled studycomparing different emission spectra. Photodermatol Photoimmunol Photomed8:73±78, 1991

de Saizieu A, Certa U, Warrington J, Gray C, Keck W, Mous J: Bacterial transcriptimaging by hybridization of total RNA to oligonucleotide arrays [seecomments]. Nat Biotechnol 16:45±48, 1998

Swensson O, Langbein L, McMillan JR, et al: Specialized keratin expression patternin human ridged skin as an adaptation to high physical stress. Br J Dermatol139:767±775, 1998

Trautinger F, Kindas-Mugge I, Dekrout B, Knobler RM, Metze D: Expression ofthe 27-kDa heat shock protein in human epidermis and in epidermalneoplasms: an immunohistological study. Br J Dermatol 133:194±202, 1995

Vogt T, Mathieu-Daude F, Kullmann F, Welsh J, McClelland M (eds): ArbitrarilyPrimed PCR of DNA and RNA. New York: John Wiley. & Sons, Inc., 1997)

Wang Y, Rea T, Bian J, Gray S, Sun Y: Identi®cation of the genes responsive toetoposide-induced apoptosis: application of DNA chip technology. FEBS Lett445:269±273, 1999

Zelicof A, Protopopov V, David D, Lin XY, Lustgarten V, Gerst JE: Two separatefunctions are encoded by the carboxyl-terminal domains of the yeast cyclase-associated protein and its mammalian homologs. Dimerization and actinbinding. J Biol Chem 271:18243±18252, 1996

988 BECKER ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY